Analysis And Application Of Probability Model Of Intravascular Ultrasound Speckle

Ultrasonic imaging makes use of ultrasonic echo signals during the process of ultrasonic propagation in the tissue.In the process of ultrasonic propagation,a large number of random scattered particles are formed due to the inhomogeneity of acoustic impedance and the randomness of spatial distribution in soft tissue of human body.The scattering waves result in the speckle pattern in ultrasonic image because of the interference and destructive interference during its propagation.The characteristic of the speckle statistical distribution is closely related to the number of scattered particles and the intensity of scattering,and directly associated with the anatomical structure characteristics of human organs.Therefore,the physical characteristics of the speckles reflect the structural characteristics of the tissues.According to the number and nature of the scattered particles in an ultrasonic imaging unit,the speckles can be divided into the following types:fully developed,fully solved,partially developed and partially solved.There are many techniques to describe speckles and applications on it,mainly in four aspects:statistics based,model based,signal processing and geometric methods.Among them,the statistical method has gradually aroused people’s attention because of its simplicity and effectiveness,and has been widely developed.The speckle distribution model has been studied from radio frequency signal and envelope signal(ultrasonic image).Through the parameter estimation of the statistical model,including the maximum likelihood estimation,the maximum posteriori estimation and the EM algorithm,the tissue characteristics of each organ are obtained.Speckles have statistical distribution characteristics,and the speckles produced by plaques of IVUS are fully developed.Based on fully developed speckles,3 kinds of models have been proposed:Rayleigh Mixture Model,Nakagami Mixture Model and Gamma Mixture Model.The above study suggests that speckles obey a single type of distribution,and points out that the probability distribution of the speckles are best described by Gamma Mixture Model.On the basis of its research,this paper used Gaussian Mixture Model and Gamma Mixture Model to describe plaques and normal vascular tissues in IVUS images respectively,and studied the influence of the number of different probability components.Finally,three evaluation indexes such as KS,KL test and correlation coefficient,are used to evaluate the experimental results,and found that the probability distribution of speckles in calcified plaques and normal vascular areas are better fitted by Gaussian Mixture Model,whereas the soft plaques are better fitted by Gamma Mixture Model.It shows that for different organizations,Gaussian Mixture Model and Gamma Mixture Model can be chosen to describe them respectively.Using the study above,the Gaussian and Gamma Mixture Model are proposed.And a probability mixture model with neighborhood smoothing is proposed.Using the maximum posterior probability criterion,the image segmentation of IVUS images is realized.The superiority of this method is proved by the comparison of other methods of mixture model for image segmentation.